DIPS, LIPS, and Hipsby David Gassko
December 28, 2006
Thank Jay Jaffe for this column. Or, alternately, send your hate mail his way—either way, all that follows is due to him.
See, Jay e-mailed me about a week ago asking for some DIPS 3.0 numbers. I had actually calculated everything at the end of the season, but had forgotten about it. Well, no more. We’re about to embark on a journey of finding the luckiest and unluckiest pitchers of 2006, based on the batted ball data that we publish (and some that we don’t) here at the Hardball Times.
But first, a dilemma. The system I use has undergone some revisions since I first introduced it, and not everyone is happy about them. As I found in both last year’s Annual and this year’s edition (if you have yet to, buy it now!), pitchers have little control over their line drive rates as well as the rate of home runs they allow per outfield fly. Sure, they have some control, but not much.
When I released last year’s numbers, I corrected for that by substituting a league average line drive rate for every pitcher as well as a league average home run per outfield fly rate. Not everyone was happy. For while that statistic may have given us a more accurate measure of a pitcher’s ability, it was going against the very basic idea behind DIPS, which was to account for those statistics which are impacted by only the pitcher and the batter, and to ignore those which may be impacted by fielders.
Upon further review, I believe those critics were right, so this year I am going back to using actual line drive and home run numbers. However, because I don’t think that taking out all the numbers that are heavily dependent on luck is a foolish idea, either, I am also going to track what I am going to call LIPS, for Luck Independent Pitching Statistics.
A quick recap of how I derive each statistic (you can skip this paragraph if you’d like). I take each player’s batted ball line, and based on league averages, I convert it into expected single, doubles, triples, home runs, reached on error, outs, and grounded into double plays. For DIPS 3.0, I use actual home runs instead of expected. For LIPS, I first transform the player’s batted ball line by substituting a league average line drive rate. Then I add in the pitcher’s actual walk, strikeout, and hit batter numbers, and calculate his expected run average (as opposed to earned run average) using BaseRuns .
Essentially, the idea behind both DIPS 3.0 and LIPS is that a lot of luck goes into a pitcher’s performance, as measured by ERA, every year. To get a better feel for how good a pitcher actually was, we want to remove the elements of his pitching line that could be especially overwhelmed by luck. It so happens that while defense independent events like strikeouts, walks, and hit-by-pitch are pretty stable from year-to-year, things like singles per outfield fly ball or line drives are not. DIPS and LIPS try to correct for that.
So who were the luckiest pitchers (based on runs above replacement) in 2006?
NAME LIPS RA RA C. Wang 5.64 3.80 J. Jennings 5.23 3.99 B. Zito 5.21 4.03 K. Rogers 5.49 4.28 A. Sanchez 5.02 3.07 A. Cook 5.55 4.53 C. Hensley 5.00 3.95 C. Carpenter 4.07 3.29 R. Halladay 4.11 3.55 J. Verlander 4.73 3.77 NAME DIPS RA RA C. Wang 4.95 3.80 J. Verlander 5.11 3.77 R. Halladay 4.32 3.55 K. Rogers 5.31 4.28 K. Saarloos 6.96 5.19 B. Arroyo 4.51 3.66 N. Robertson 5.20 4.23 J. Suppan 5.78 4.74 T. Glavine 5.17 4.27 C. Carpenter 4.00 3.29The fact that Chien-Ming Wang is at the top of both lists is not a good thing for Yankees fans. What’s worse is that I trust LIPS even more than I do DIPS 3.0. While Wang was worth 52 runs above replacement last year, he should have been worth about seven, according to LIPS, so Yankees fans should expect a big step backwards this year. With the Yankees offense, Wang should still rack up wins, but of course, so could I.
Chris Carpenter and Roy Halladay are players I would worry less about because even their DIPS/LIPS numbers are very good. I would not expect either to put up an ERA below 3.00 next year, but both should stay in the mid-to-high threes, which is still quite good.
The presence of two Tigers on each of the lists is at least partially attributable to the Tigers defense overall; last year, Tigers pitchers allowed 53 less singles and 27 less doubles than expected (though they did allow 13 more triples).
And of course, their DIPS/LIPS numbers should give some pause to teams pursuing Barry Zito and Jeff Suppan. Zito’s DIPS 3.0 RA is 4.65, which is above average, but probably not worth $100 million, while Suppan’s LIPS RA is 5.29, which is well below average (as is, of course, his 5.78 DIPS 3.0 RA). The Mets may regret the “bargain” that was Tom Glavine as well.
But what about the other side of the coin? Who were the unluckiest pitchers of 2006?
NAME LIPS RA RA R. Lopez 4.95 6.14 R. Johnson 4.39 5.49 D. Turnbow 4.33 8.15 B. Chen 5.25 7.39 J. Towers 5.65 9.00 K. Davies 5.39 8.53 J. Vazquez 4.24 5.15 G. Rusch 4.90 7.73 S. Baker 4.68 6.80 T. Buchholz 4.75 6.37 NAME DIPS RA RA R. Johnson 4.13 5.49 J. Vazquez 4.10 5.15 M. Cain 3.52 4.39 J. Benoit 3.40 5.54 S. Baker 5.01 6.80 B. Sheets 2.68 3.99 D. Turnbow 5.39 8.15 P. Astacio 4.92 6.38 O. Perez 5.20 7.43 M. Clement 4.99 6.89
But where Wang’s DIPS/LIPS numbers are bad news for the Yankees, Randy Johnson’s are great news. New York can legitimately expect an under-4.00 ERA from the Big Unit next season, which will make for a nice comeback after his 5.00 ERA in ‘06.
Two possible 2007 Cy Young contenders show up on the DIPS 3.0 list, as Ben Sheets puts up especially strong numbers and Matt Cain looks ready to burst onto the scene at just 22.
Overall, what we see here is a list of decent pitchers who had (undeservedly) terrible years in 2006. Taylor Buchholz, Derrick Turnbow, Matt Clement—none of these guys are going to be aces next year, but they’re not going to post RAs over 6.00 again either.
One of the big differences between DIPS 3.0 and LIPS is that LIPS substitutes a league average home run rate based on the number of line drives and fly balls allowed by a pitcher, while DIPS uses his actual home runs. What pitchers does this make the biggest difference for?
The ten luckiest pitchers in 2006 according to LIPS, based on home runs expected NAME lHR HR J. Lackey 27 14 J. Blanton 27 17 J. Wright 19 10 J. Jennings 26 17 J. Schmidt 30 21 E. Santana 29 21 M. Cain 25 18 R. Lugo 11 4 Z. Duke 23 17 K. Escobar 23 17 The ten unluckiest pitchers in 2006 according to LIPS, based on home runs expected NAME lHR HR J. Beckett 25 36 C. Silva 27 38 G. Rusch 10 21 B. Chen 18 28 M. Mulder 10 19 M. Buehrle 27 36 R. Ortiz 7 15 F. Hernandez 15 23 K. Benson 26 33 A. Burgos 9 16
Some of these numbers are staggering. Greg Rybarczyk wrote a great article discussing Josh Beckett’s home run totals in the 2007 edition of the Hardball Times Annual, and the LIPS stats show that Beckett allowed many more home runs than we would have expected. Expect him to have a much better season next year.
As I found in the Annual, there is some year-to-year consistency is home run per fly ball numbers (though not much), so I would not expect these guys to rebound totally to their expected home run numbers, but they should be much, much closer next year. Mark Mulder especially looks like a guy who just got hit hard because of injuries, and should be much better come 2007.
Seeing two Giants on one list and three Orioles on the other might make you a bit suspicious. For the Giants, there is due cause—their stadium suppresses home runs to the tune of four less home runs per thousand outfield flies. So while both Jason Schmidt and Cain were still quite lucky, part of their luck is due to AT&T Park. The Orioles, on the other hand, play in a perfectly neutral park—they were just extremely unlucky.
So what can we learn from exercises like this? Well, DIPS 3.0, in its current incarnation, does a better job of separating a pitcher from his defense than any previous version of DIPS. LIPS surpasses any previous statistic in separating the repeatable part of a pitcher’s performance from luck. Both can help illuminate and explain past seasons, and give us a better handle on what to expect in the future.
Full DIPS 3.0 and LIPS statistics can be accessed here.
References and Resources
Note: The file linked to above has two tabs; the first is DIPS 3.0 numbers, the second LIPS. The bolded categories are "DIPS numbers," that is the numbers the pitcher would be expected to have given his batted ball distribution. The non-bolded numbers are his actual statistics. I've calculated runs above replacement for each pitcher, both using his actual numbers and his DIPS/LIPS numbers (and in fact, the first two lists in this article were ordered based on the differences in runs above replacement). I think this is a great resource, and I hope you can enjoy it.
David Gassko is a former consultant to a major league team. He welcomes comments via e-mail.